Re: Lilypond error behaviour

[Noeck]

Am 17.04.2016 um 16:49 schrieb Kieren MacMillan:
>> Except that the premise of this thread was that users […]
>> > want to be able to deduce the presence of errors from the existence of 
>> > output files.
> Ah… sorry, I didn’t get that that was the premise.

I don't think so. To me it seems quite the contrary, the users wanted to
deduce from the error code whether Lilypond could do something with the
input to produce the output and *not* have to figure that out from the
presence of the output file. But that's just what I read.


>From my point of view, the discussion boils down to:

- Lilypond has two categories of 'problems': warnings and errors and the
developers categorize the situation according to the associated severity.
-> metric: severe or not?

- Another criterion implied in this thread would be:
Can Lilypond go on after this situation and try to finish whatever it
can or is there no chance and it just stops here before it reaches the
end of the task.
-> metric: finishable (with some output) or exit here?


These two are often similar but not always.
IMHO, it would make sense to consider both and have 3 categories:
- warning: user, please look at this
- error: this is severe, there is something definitely wrong, but
  Lilypond did its very best to keep running
- fatal error: this is severe and Lilypond could not rescue the
  situation, nothing was produced

Probably, some similar situations (similar input syntax error) appear
inconsistent, because it depends on what is called later on whether this
is fatal or not. But still but the severity would be consistent (warning
or error) and the pdf would be there if there is no fatal error.

Cheers,
Joram